Identifying and patching vulnerabilities of camera-LiDAR based Autonomous Driving Systems

The rise of autonomous vehicles (AVs) is transforming the transportation sector, potentially enhancing road safety, optimizing traffic flow, and bringing about a more sustainable future. Central to this revolution lies two interlinked technological keystones: integrating advanced sensor systems and applying cutting-edge machine-learning techniques. Specifically, the fusion of high-resolution imagery from cameras and the depth precision of Light Detection and Ranging (LiDAR) sensors equips AVs with an unparalleled perceptual prowess, allowing AVs to capture a holistic, 360-degree spatial awareness of their surroundings. Subsequently, machine learning algorithms transform the collected sensor data into actionable insights, empowering the vehicle to make accurate and informed driving decisions. While machine learning algorithms help autonomous driving systems exhibit remarkable capabilities in recognizing patterns and making decisions, they also harbor an Achilles' heel known as adversarial vulnerability. It has been previously shown that attacks can mislead the vehicle into misrecognizing traffic signs, misjudging obstacles, or misinterpreting road conditions. Such vulnerabilities pose profound safety risks, as malicious actors could exploit them to induce unintended behaviors in AVs, potentially leading to hazardous situations on the road. As self-driving technology accelerates, understanding and mitigating these adversarial vulnerabilities becomes paramount to ensure the safety, reliability, and public trust in autonomous transportation. This project aims to provide a multi-dimensional security analysis for advanced autonomous driving systems. Specifically, the research team pivots their investigation toward the Bird's Eye View (BEV) — a cutting-edge 3D perception system now gaining traction in real-world self-driving systems. The perceptual capabilities of this considered system will be further enhanced via the integration with LiDAR signal. It is noteworthy that despite its growing prevalence in modern AVs, the BEV system remains a relatively untapped area in adversarial machine learning research. Moreover, beyond merely focusing on fooling AVs’ perception system to recognize objects of interest as in existing studies wrongly, this project orients towards adversarial scenarios where attackers can induce tangible, real-world disruptions — such as instigating traffic congestions or triggering vehicular collisions — especially when interacting with other dynamic agents like vehicles or pedestrians.

• Record URL:
  • https://www.clemson.edu/cecas/tracr/research/projects.html

Language

• English

Project

• Status: Active
• Funding: $112393
• Contract Numbers:

  69A3552344812

  69A3552348317

• Sponsor Organizations:

  Office of the Assistant Secretary for Research and Technology

  University Transportation Centers Program
  Department of Transportation
  Washington, DC  United States  20590

  University of California, Santa Cruz

  1156 High Street, Mail Stop SOE2
  Santa Cruz, California  United Kingdom  95064

  University of Texas at Dallas

  800 W Campbell Rd
  Richardson, Texas  United States  75080
• Managing Organizations:

  National Center for Transportation Cybersecurity and Resiliency (TraCR)

  Clemson University
  Clemson, SC  United States 

  University of California, Santa Cruz

  1156 High Street, Mail Stop SOE2
  Santa Cruz, California  United Kingdom  95064
• Project Managers:

  Chowdhury, Mashrur

• Performing Organizations:

  University of California, Santa Cruz

  1156 High Street, Mail Stop SOE2
  Santa Cruz, California  United Kingdom  95064

  University of Texas at Dallas

  800 W Campbell Rd
  Richardson, Texas  United States  75080
• Principal Investigators:

  Xie, Cihang

  Cardenas, Alvaro

  Kantarcioglu, Murat

• Start Date: 20240101
• Expected Completion Date: 20241231
• Actual Completion Date: 0
• USDOT Program: University Transportation Centers

Subject/Index Terms

• TRT Terms: Autonomous vehicles; Cameras; Computer security; Laser radar; Machine learning; Physical phenomena; Safety; Sensors
• Subject Areas: Data and Information Technology; Planning and Forecasting; Safety and Human Factors; Security and Emergencies; Transportation (General); Vehicles and Equipment;

Filing Info

• Accession Number: 01906996
• Record Type: Research project
• Source Agency: National Center for Transportation Cybersecurity and Resiliency (TraCR)
• Contract Numbers: 69A3552344812, 69A3552348317
• Files: UTC, RIP
• Created Date: Feb 5 2024 4:04PM